Stroke is a leading cause of disability, being motor impairments its major consequence. Despite rehabilitation, more than 60% of the patients report upper-limb motor dysfunction. The design of novel rehabilitation strategies requires objective measures to assess motor impairment and recovery. In a previous study, we proposed to use the synergy components of the unaffected limb as a reference to be targeted by rehabilitation, since they are proven to explain healthy motor control and to be altered after stroke. We demonstrated that healthy subjects have very similar control structures (synergies and activation vectors) in their right and left arms. Here, we investigate the existence of movement-specific control strategies. To do so, we analyze the inter-subject similarity of the healthy control structure in twelve common stroke rehabilitation exercises and we evidence that motor control is movement specific and generalizes across different subjects and their limbs. However, the similarity degree depends on the movement, suggesting that novel training protocols should purposely choose the rehabilitation exercises to ensure maximum control similarity with the reference pattern.